This invention relates to the manufacturing of semiconductor lasers and, more particularly, to a reliability assurance technique for such lasers.
The reliability of a lightwave system is a function of the reliability of the components which make up that system. The manufacturer of the system establishes quality standards for the components in such a way that overall system performance is assured within statistical confidence limits. The standards vary, however, depending on the system user's tolerance for system failure; i.e., the user's ability to absorb the costs of system down-time and component repair or replacement. In a terrestrial system, for example, it may be relatively easy to replace a failed component in a manhole, but in a submarine cable system it is extremely difficult and expensive to raise the cable from the ocean floor to the surface to effect repairs. As a consequence redundancy is built into such systems and lifetime requirements may exceed 25 years.
Long-lifetime requirements imply stringent quality assurance standards. Components are subjected to a regimen of electrical, optical and mechanical screening procedures in order to warrant the reliability of a given population of devices. For example, critical lightwave components such as laser diodes are visually inspected to uncover visible flaws (e.g., morphological defects), mechanically stressed to test chip and wire bonds, and thermally cycled to uncover cracks or flaws in the semiconductor material. In addition, light-current (L-I) and current-voltage (I-V) characteristics (and their derivatives) are measurd to determine threshold current, series resistance and the like. Finally, the lasers are subjected to accelerated aging, during which the lasers are operated for 1000s of hours at a typical optical output power (e.g., 3-5 mW/facet) at a higher-than-normal temperature (e.g., 60.degree. C.). As the output power tends to decrease with time, a feedback control circuit increases the drive current in order to maintain the output power at a constant level. When the drive current exceeds a predetermined limit (e.g., a 50% increase), a laser is said to have failed. Performing accelerated aging on numerous lasers establishes a measure of reliability and may suggest design changes to enhance device lifetime.
Once the design is fixed, however, this type of testing does not insure reliability of a given population of lasers because while some laser failure modes are strongly temperature activated, others have little or no temperature dependence. For example, dark line defects (DLDs) are weakly temperature dependent. On the other hand, long term degradation in lasers is strongly temperature dependent.
Long-term degradation is the popular failure mode to address. Most prior art work focuses on it. Although it is strongly temperature activated, it may not be possible to operate the devices at high temperatures, D. S. Peck and C. H. Zierdt, IEEE Proceedings, Vol. 62, No. 2, pp. 185-211 (1974). As a result, degradation studies are usually lengthy and can be confusing because early degradation behavior may exhibit transient modes until the devices stabilize. Consequently, rates and acceleration factors (or the associated equivalent activation energies) are not always credibly determined. The observed degradation (increase in operating current at constant output power) of lasers of the etched mesa type operating at 1.3 .mu.m may be used to illustrate the transient or saturable mode. These lasers are described by M. Hirao et al, Journal of Applied Physics Vol. 51, p. 4539 (1980).
Shown in FIG. 1 herein are a number of semiconductor laser degradation patterns for operating current which produces an optical power of 3 mW/facet CW in a 60.degree. C. ambient. These lasers had previously shown less than a 5% increase in operating current after a burn-in of 100 hours in a 60.degree. C. ambient at an optical power of 5 mW/facet. Some lasers show a constant rate of increase; others show highly variable rates. Despite appearances, all of the lasers are exhibiting a transient behavior mode and some have not yet stabilized. As a result, the long-term rate of aging is not known. Moreover, step-stress aging at different temperatures yields acceleration factors or equivalent activation energies that depend on the direction of the step. Consequently, predictions based on either isothermal or step stress aging may be inaccurate unless the aging rate behavior is stabilized.
Study of sudden failure modes by high temperature, bias aging is another typical reliability assurance technique as discussed by D. S. Peck et al supra. However, it frequently provides no information about failure modes that are, at best, weakly temperature activated. These modes represent a nightmare for the device designer. Too often these failures begin to show up late in the qualification cycle, sometimes long after early field deployment. See, for example, W. C. Ballamy et al IEEE Transactions on Electron Devices, Vol. ED-25, p. 746 (1978) or A. S. Jordan et al, 18th Annual Proceedings Reliability Physics Symposium, Las Vegas, p. 123 (1980).